The present invention relates to tape storage systems, and more specifically, to selectively applying erasure decoding on a per-track basis.
Currently-used linear tape drives apply product codes for error-correction coding (ECC). These product codes contain two Reed-Solomon component codes consisting of a C1 row code and a C2 column code. Failure to decode a product codeword, which requires successful decoding of all C1 rows and all C2 columns within a product code, leads to a temporary and/or permanent error. These temporary or permanent errors are a significant problem when attempting to store data to tape.
Typically, current tape drives implement ECC with two modes of operation: 1) Error-only decoding (decoding in which no information about uncorrectable C1 codewords is passed to the C2 decoder); and 2) Erasure-decoding (decoding in which information about C1 uncorrectable codewords are passed from the C1 decoder to the C2 decoder).
The second mode, erasure-decoding, is beneficial in the case where C1 uncorrectable codewords having a large number of byte errors are used as erasure flags for C2 decoding. Conversely, the second mode likely leads to performance degradation in the case where C1 uncorrectables with only a small number of errors (only slightly beyond the error correction capability of the C1 decoder) are used as erasure flags for C2 decoding.
Therefore, it is not always beneficial to utilize erasure-decoding when implementing ECC in tape drives.